Investigating spatial variation in the relationships between NDVI and environmental factors at multi-scales: a case study of Guizhou Karst Plateau,China

Knowing the spatial relationships between the normalized difference vegetation index (NDVI) and environmental variables is of great importance for monitoring rocky desertification. This article investigated the spatially non-stationary relationships between NDVI and environmental factors using geographically weighted regression (GWR) at multi-scales. The spatial scale-dependency of the relationships between NDVI and environmental factors was identified by scaling the bandwidth of the GWR model, and the appropriate bandwidth of the GWR model for each variable was determined. All GWR models represented significant improvements of model performance over their corresponding ordinary least squares (OLS) models. GWR models also successfully reduced the spatial autocorrelations of residuals. The spatial relationships between NDVI and environmental factors significantly varied over space, and clear spatial patterns of slope parameters and local coefficient of determination (R ²) were found from the results of the GWR models. The study revealed detailed site information on the different roles of related factors in different parts of the study area, and thus improved the model ability to explain the local situation of NDVI.     

Spatial non-stationarity and scale-dependency of prediction accuracy in the remote estimation of LAI over a tropical rainforest in Sulawesi, Indonesia

The technique of Geographically Weighted Regression (GWR) was used for estimation of Leaf Area Index (LAI) from remote sensing-based multi-spectral vegetation indices (VI) such as Normalized Difference Vegetation Index (NDVI), the mid-infrared corrected Normalized Difference Vegetation Index (NDVIc), Simple Ratio (SR), Soil-Adjusted Vegetation Index (SAVI) and Reduced Simple Ratio (RSR) in a region of equatorial rainforest in Central Sulavesi, Indonesia. The linear regressions between NDVI, NDVIc, SR, SAVI and RSR as explanatory variables and ground measurements of LAI at 166 plots as a dependent variable were produced using common modelling approach — Ordinary Least Squares (OLS) regression fitted to all data points, as well as GWR. Accuracy and precision statistics indicate that the GWR method made significantly better predictions of LAI in all simulations than OLS did. The relationships between LAI and the explanatory variables were found to be significantly spatially variable and scale-dependent. GWR has the potential to reveal local patterns in the spatial distribution of parameter estimates, it demonstrated sensitivity of the model's accuracy and performance to scale variation. The GWR approach enables finding the most appropriate scale for data analysis. This scale was different for each VI. The results suggest that spatial non-stationarity and scale-dependency in the relationship between LAI and remote sensing data has important implications for estimations of LAI based on empirical transfer functions.     

The influence of sampling density on geographically weighted regression: a case study using forest canopy height and optical data

Geographically weighted regression (GWR) extends the conventional ordinary least squares (OLS) regression technique by considering spatial nonstationarity in variable relationships and allowing the use of spatially varying coefficients in linear models. Previous forest studies have demonstrated the better performance of GWR compared to OLS when calibrated and validated at sampled locations where field measurements are collected. However, the use of GWR for remote-sensing applications requires generating estimates and evaluating the model performance for the large image scene, not just for sampled locations. In this study, we introduce GWR to estimate forest canopy height using high spatial resolution Quickbird (QB) imagery and evaluate the influence of sampling density on GWR. We also examine four commonly used spatial analysis techniques – OLS, inverse distance weighting (IDW), ordinary kriging (OK) and cokriging (COK) – and compare their performance with that using GWR. Results show that (i) GWR outperformed OLS at all sampling densities; however, they produced similar results at low sampling densities, suggesting that GWR may not produce significantly better results than OLS in remote-sensing operational applications where only a small number of field data are collected. (ii) The performance of GWR was better than those of IDW, OK and COK at most sampling densities. Among the spatial interpolation techniques we examined, IDW was the best to estimate the canopy height at most densities, while COK outperformed OK only marginally and produced larger canopy height estimation errors than both IDW and GWR. (iii) GWR had the advantage of generating canopy height estimation maps with more accurate estimates than OLS, and it preserved patterns of geographic features better than IDW, OK or COK.     

Sensing urban poverty: From the perspective of human perception-based greenery and open-space landscapes

Greenery and open spaces play significant roles in environmentally sustainable societies, providing urban ecosystem services and economic benefits that reduce urban poverty. Current urban poverty research has solely focused on top-down observations or direct human exposure to greenery and open spaces and has failed to sense landscape characteristics, including occupation and inequality, representing the social attributes of urban poverty. This paper demonstrates the potential to better understand certain social characteristics, including occupation and inequality between urban greenery and open spaces, and to further investigate their relationship with urban poverty. Percentage and aggregation indicators are proposed based on street view images to estimate the occupation and inequality between human perception-based greenery and open spaces. The relationship between human perception and urban poverty is accordingly analysed using geographically weighted regression (GWR). The GWR model results attain an R-squared value of 0.583 and further reveal that the relationships between human perception-based landscapes and urban poverty are spatially non-stationary, indicating varying relationships across space. This implication leads to an improved understanding of the relationship between greenery and open-space landscapes and living conditions and to further allowing effective policies to help identify deprived areas.     

Establishing the relationship between urban land-cover configuration and night time land-surface temperature using spatial regression

Studies suggest that urban form can influence microclimate regulation. Remote sensing studies have contributed to these findings through analysis of high-resolution land cover maps, landscape ecology metrics, and thermal imagery. Collectively, these have been referred to as land cover configuration studies. There are three objectives to this study. The first is to assess the relationship between nighttime land surface temperatures (LST) and land cover configuration and composition. The second objective is to outline a comprehensive methodology that includes ordinary least squares (OLS), spatial regression, variable selection, and multicollinearity analysis. Our last objective is to test three hypotheses about the relationship between LST and land cover, which can briefly be described as: 1) the importance of land-use regimes in modeling LST from land cover composition and configuration variables; 2) the strength of the correlation between LST and roads, buildings, and vegetation; and 3) the improved quality of models using landscape metrics in modeling the relationship between LST and land cover. Based on 16 different models (8 OLS, 8 spatial regression) we could confirm the above hypotheses, but we found that the configuration of buildings, roads, and vegetation have a complex relationship with LST. Our interpretation of this complexity, combined with the strength of composition variables, is that parsimonious models, for now, are more useful to urban planners because they are more generalizable. Finally, spatial regression models of land cover configuration and LST demonstrated an improvement over non-spatial linear models (OLS). Spatial regression models reduced heteroskedasticity and clusters of residuals, and tempered coefficients, suggesting that the OLS models could be biased. OLS models were still found to be a valuable tool for exploratory analysis.     

Modelling tree diameter from airborne laser scanning derived variables: A comparison of spatial statistical models

Formerly, tree height has been more difficult to measure accurately in the field than tree diameter at breast height. As a consequence, models to predict height from diameter measurements have been widely developed in the forestry literature. Through the use of airborne laser scanning technology (e.g., LiDAR), tree variables such as height and crown diameter can be measured accurately, a development which has spawned the need for models to predict diameter from airborne laser-derived measurements. Although some work has been done for fitting such models, none have incorporated spatial information to improve the accuracy of the predicted diameters. Using a simple linear model for predicting tree diameter from laser-derived tree height and crown diameter measurements, we compared the performance of ordinary least squares (OLS), generalized least squares with a non-null correlation structure (GLS), linear mixed-effects model (LME), and geographically weighted regression (GWR). Our data were obtained from 36 sample plots established in Norway. This is the first study to examine the use of spatial statistical models for tree-level LiDAR data. Root mean square prediction errors in tree diameter with LME are 3.5%, with GWR are 10%, and with OLS and GLS are 17%. LME also exhibited low variability in predicting performance across all the validation classes (based on laser-derived height). Giving the difficulties of using parametric statistical inference (such as maximum likelihood-based indices) for GWR, we used permutation tests as a way for detecting statistical differences. LME was significantly better than the other models, as well as GWR was to OLS and GLS. Our results indicate that the LME model produced the best predictions of tree diameter from LiDAR-based variables to a degree that has previously not been possible.     

基于城镇居民用地再分类的人口数据空间化方法研究——以长江中游4省为例

基于土地利用数据的人口统计数据空间化方法,在处理过程中会出现同一土地利用类型下人口难以细分的情况,从而影响人口空间数据精度。引入夜间灯光信息并提出了一种基于夜间灯光强度对城镇居民地再分类的人口空间化方法,以改善人口空间数据精度。基于DMSP/OLS夜间灯光及土地利用数据,以长江中游4省为研究区进行方法试验。研究结果显示:利用夜间灯光数据对城镇居民地再分类后,各分区模型的调整R2都提高到了0.8以上,人口空间数据总体平均相对误差较重分类前降低了12.32%。说明该方法在提高传统人口数据空间化模型精度的基础上能够细化城镇居民地人口空间分布。     

The relationship between centrality and land use patterns: Empirical evidence from five Chinese metropolises

The relationship between location and land use patterns is one of the classic theoretical issues in urban studies. Classic models based on the monocentricity hypothesis have limitations in the interpretation of modern urban structure. China has experienced institutional transformation in recent decades, and the interaction of national government policy, market forces and the natural environment has influenced urban planning in Chinese metropolises, resulting in urban structures with special characteristics. This paper examines the distribution of location and land use intensity, and tested the Alonso model by the relationship between them in five Chinese metropolises using Point of Interest data, space syntax methodology, the grid weighted statistical method and the Geographically Weighted Regression (GWR) model. Universal patterns about the scaling relation between intensity of land use types and the centrality of location are revealed. The elasticity of land use types to location, from high to low sensitivity, is commercial, residential then industrial land in most of the five metropolises studied. The sensitivity sequence of land use studied suggests that the hypothetical model based on the classical Alonso model can explain the spatial structure of modern metropolises in China to some extent, especially for the commercial land. But the order of sensitivity of residential land and industrial land to location does not conform to the model. The spatial heterogeneity in land use intensity and centrality were explored and the factors embedded were discussed. It can be found that the relation between centrality and land use intensity conforms to power law. In most of the metropolises studied, when the scaling relation between land use intensity and centrality is super linear, the sequence of the frequency value from high to low are commercial, residential and industrial land; when the scaling relation is sublinear, the sequence of the frequency value is industrial, residential and commercial land.     

The AMSL LST algorithm validated for the Xinjiang Autonomous Region in China

The split-window land surface brightness temperature (LST_b) algorithm of Coll and Caselles (1994) is one of the first approaches to estimate LST_b applied for large surface areas. In this article, we describe a calibrated and validated version of the Coll and Caselles (1994) algorithm applied for the retrieval of land surface air temperature (LST_a) – equivalent to standard WMO (World Meteorological Organization) temperature measurements – for the province of Xinjiang (PR of China). Locally received MODIS (Moderate Resolution Imaging Spectroradiometer) imagery (Fukang receiving station) is used as the input data stream for the so-called AMSL (Aqua MODIS SWA LST_a) algorithm. The objective to develop this algorithm is that it is an input for a distributed hydrological model as well as a soil moisture content retrieval algorithm. In the Xinjiang province with an abundance of arid to semi-arid regions, a highly continental climate, irrigated crop fields and mountain ranges of 6000 m and higher, one typically deals with the spatio-temporally complex conditions, making a high-accuracy retrieval of LST_a quite a challenge. The calibration and validation of the AMSL LST_a product (LST_a,amsl) – using the Jackknife method – is performed using LST_a measurements (LST_a,tmb) from 49 meteorological stations managed by the Tarim Meteorological Bureau (TMB). These stations are distributed relatively homogeneously over the province. The TMB stations’ temperature data are split into 40 calibration LST_a,tmb data sets and 9 validation LST_a,tmb data sets. We can observe that when validated, the LST_a,amsl versus LST_a,tmb validation relationship elicits a high correlation, a slope very close to 1 and an intercept very close to 0. The validated LST_a,amsl estimates demonstrate an estimation accuracy of 0.5 K. The results presented in this article suggest that the LST_a,amsl product is suitable to estimate the land surface air temperature spatio-temporal fields for the arid and semi-arid regions of the Xinjiang province accurately.     

Soil moisture estimation over grass-covered areas using AIRSAR

An empirical study of the relation between the AIRSAR's signals and land surface parameters is conducted using data collected during MACEUROPE' 91. General additive regression models are fitted to the data simulated from a calibrated microwave backscattering model. A comparison between the model predictions and field observations indicates that the regression models are good predictors to the AIRSAR's signals over the grass-covered areas. Based on the regression relationships, a soil moisture retrieval algorithm combining the Lvband multi-polarization AIRSAR data is proposed and used to create spatial soil moisture maps of the Slapton Wood catchment.     

Modeling urban growth using spatially heterogeneous cellular automata models: Comparison of spatial lag,spatial error and GWR

Many methods can be used to construct geographical cellular automata (CA) models of urban land use, but most do not adequately capture spatial heterogeneity in urban dynamics. Spatial regression is particularly appropriate to address the problem to reproduce urban patterns. To examine the advantages and disadvantages of spatial regression, we compare a spatial lag CA model (SLM-CA), a spatial error CA model (SEM-CA) and a geographically-weighted regression CA model (GWR-CA) by simulating urban growth at Nanjing, China. Each CA model is calibrated from 1995 to 2005 and validated from 2005 to 2015. Among these, SLM and SEM are spatial autoregressive (SAR) models that consider spatial autocorrelation of urban growth and yield highly similar land transition probability maps. Both SAR-CA and GWR-CA accurately reproduce urban growth at Nanjing during the calibration and validation phases, yielding overall accuracies (OAs) exceeding 94% and 85%, respectively. SAR-CA is superior in simulating urban growth when measured by OA and figure-of-merit (FOM) while GWR-CA is superior regarding the ability to address spatial heterogeneity. A concentric ring buffer-based assessment shows OA valleys that correspond to FOM peaks, where the ranges of valleys and peaks indicate the areas with active urban development. By comparison, SAR-CA captures more newly-urbanized patches in highly-dense urban areas and shows better performance in terms of simulation accuracy; whereas, GWR-CA captures more in the suburbs and shows better ability to address spatial heterogeneity. Our results demonstrate that spatial regression can help produce accurate simulations of urban dynamics featured by spatial heterogeneity, either implicitly or explicitly. Our work should help select appropriate CA models of urban growth in different terrain and socioeconomic contexts.     

Assessing diel urban climate dynamics using a land surface temperature harmonization model

The main objective of this study is to examine how climate gradients (coastal to inland climate) and land-cover types affect land surface temperature (LST) diel variation. To achieve this, we applied LST harmonization model, which integrates LST at daytime and night-time using sine and cosine functions, to reconstruct a complete diel LST curve for census block groups (CBGs) with both highly vegetated and impervious land-cover types in 10 major cities of the Los Angeles region distributed throughout the coastal to inland climate gradient. We calculated diel LST metrics of minimum LST (LST_min), maximum LST (LST_max), diel LST range (DLSTR), and time of LST_min and LST_max for each CBG as well as LST differences between neighborhoods with extensive (>80%) impervious and vegetated surface. First, we examined how distance from coast explained the calculated LST products. Results showed that DLSTR (by factor of 2.50), LST_max (by factor of 1.57), and LST differences between CBGs with extensive impervious and vegetated surfaces (by factor of 4) were higher for cities in inland compared to the coastal cities. Time of LST_max shifted by 2.50 h from the coastal cities to the midland (regions located between coastal and inland areas) and then inland. Second, we examined how distance from coast and land-cover types explained estimated LST of CBGs at 14:00. Results showed that distance from coast and land-cover types together explained 81% of LST at 14:00. Percentage of vegetation was the most significant driver to explain LST. We concluded that using seamless LST data enables us to better evaluate temporally informative metrics of LST for use in human health, resource use, and natural resource management at regional scale.     

The influence of land-cover type on the relationship between NDVI–LST and LST-Tair

Air temperature (T_2m or T_air) measurements from 20 ground weather stations in Berlin were used to estimate the relationship between air temperature and the remotely sensed land surface temperature (LST) measured by Moderate Resolution Imaging Spectroradiometer over different land-cover types (LCT). Knowing this relationship enables a better understanding of the magnitude and pattern of Urban Heat Island (UHI), by considering the contribution of land cover in the formation of UHI. In order to understand the seasonal behaviour of this relationship, the influence of the normalized difference vegetation index (NDVI) as an indicator of degree of vegetation on LST over different LCT was investigated. In order to evaluate the influence of LCT, a regression analysis between LST and NDVI was made. The results demonstrate that the slope of regression depends on the LCT. It depicts a negative correlation between LST and NDVI over all LCTs. Our analysis indicates that the strength of correlations between LST and NDVI depends on the season, time of day, and land cover. This statistical analysis can also be used to assess the variation of the LST–T_2m relationship during day- and night-time over different land covers. The results show that LST_Day and LST_Night are correlated significantly (p = 0.0001) with T_2mDay (daytime air temperature) and T_2mNight (night-time air temperature). The correlation (r) between LST_Day and T_Day is higher in cold seasons than in warm seasons. Moreover, during cold seasons over every LCT, a higher correlation was observed during daytime than during night-time. In contrast, a reverse relationship was observed during warm seasons. It was found that in most cases, during daytime and in cold seasons, LST is lower than T_2m. In warm seasons, however, a reverse relationship was observed over all land-cover types. In every season, LST_Night was lower than or close to T_2mNight.     

Estimation of air temperature and reference evapotranspiration using MODIS land surface temperature over Greece

Moderate Resolution Imaging Spectroradiometer (MODIS), land surface temperature data, during daytime (LST_day) or night-time (LST_night), were employed for predicting maximum (T_max) or minimum (T_min) air temperature measured at ground stations, respectively, in order to be used as alternative inputs in minimum data-based reference evapotranspiration (ET) models in 28 stations in Greece during the growing season (May–October). The deviations between daily LST_night and T_min were found to be small, but they were greater between LST_day and T_max. Furthermore, the temperature vegetation index (TVX) method was employed for achieving more accurate T_max values from LST_day, after determining the normalized difference vegetation index of a full canopy (NDVI_max). The TVX method was validated on ‘temporal’ basis, but when the method was tested spatially, the improvement on the T_max estimates from LST_day was not encouraging, for being used operationally over Greece. Thus, LST_day or LST_night MODIS data were used as inputs in three ET models [Hargreaves–Samani, Droogers–Allen, and Reference Evapotranspiration Model for Complex Terrains (REMCT)] and their estimations, as compared with ground-based Penman–Monteith estimates, indicated that the REMCT model achieved the most accurate ET predictions (r = 0.93, mean bias error = 0.44 mm day^–1 and root mean square error = 0.74 mm day^–1), which can allow the spatial analysis of ET at higher spatial resolutions in areas with lack of ground temperature data.     

Assessing accessibility with local coefficients for the LUTI model MARS

Accessibility-based land use and transport interaction (LUTI) models are tools for policy assessments that facilitate coherent implementation of sustainable strategic urban plans. This study aims to improve one of those LUTI models introducing the different impact of factors influencing residential and workplace choice by computing local coefficients. In particular, this research explores the methodology of integrating the public choice model into the MARS (Metropolitan Activity Relocation Simulator) model using a complex accessibility indicator. We established a new approach to input the variation of the influence of each public service across space with the use of Geographically Weighted Regression (GWR). The model update and extension of MARS are all based on the Region of Madrid, Spain. Using the new accessibility indicator yielded better results and corrected some under estimation and overestimations in the number of workplaces. The correlation using the new accessibility indicator is significantly higher than the one using the old one. The prediction using the new indicator achieves better results for the whole area whatever the zone is small or large. The analysis evidences the convenience of GIS and LUTI combination to improve model accuracy and precision. Using the new accessibility indicator based on local coefficients, MARS model fits better with the real data in respect of the distribution of workplaces and residents, which are the key representatives of the land use sub-model.     

Support vector machines for urban growth modeling

This paper presents a novel method to model urban land use conversion using support vector machines (SVMs), a new generation of machine learning algorithms used in the classification and regression domains. This method derives the relationship between rural-urban land use change and various factors, such as population, distance to road and facilities, and surrounding land use. Our study showed that SVMs are an effective approach to estimating the land use conversion model, owing to their ability to model non-linear relationships, good generalization performance, and achievement of a global and unique optimum. The rural-urban land use conversions of New Castle County, Delaware between 1984–1992, 1992–1997, and 1997–2002 were used as a case study to demonstrate the applicability of SVMs to urban expansion modeling. The performance of SVMs was also compared with a commonly used binomial logistic regression (BLR) model, and the results, in terms of the overall modeling accuracy and McNamara’s test, consistently corroborated the better performance of SVMs.     

Integrating spatial nonstationarity into SLEUTH for urban growth modeling: A case study in the Wuhan metropolitan area

Accurate forecasting of future urban land expansion can provide useful information for policy makers and urban planners to better plan for the impacts of future land use and land cover change (LULCC) on the ecosystem. However, most current studies do not emphasize spatial variations in the influence intensities of the various driving forces, resulting in unreliable predictions of future urban development. This study aimed to enhance the capability of the SLEUTH model, a cellular automaton model that is commonly used to measure and forecast urban growth and LULCC, by embedding an urban suitability surface from geographically weighted logistic regression (GWLR). Moreover, to examine the performance of the loosely-coupled GWLR-SLEUTH model, a layer with only water bodies excluded and a layer combining the former with an urban suitability surface from logistic regression (LR) were also used in SLEUTH in separate model calibrations. This study was applied to the largest metropolitan area in central China, the Wuhan metropolitan area (WMA). Results show that the integrated GWLR-SLEUTH model performed better than either the traditional SLEUTH model or the LR-SLEUTH model. Findings demonstrate that spatial nonstationarity existed in the drivers' impacts on the urban expansion in the study area and that terrain, transportation and socioeconomic factors were the major drivers of urban expansion in the study area. Finally, with the optimal calibrated parameter sets from the GWLR-SLEUTH model, an urban land forecast from 2017 to 2035 was conducted under three scenarios: 1) business as usual; 2) under future planning policy; and 3) ecologically sustainable growth. Findings show that future planning policy may promise a more sustainable urban development if the plan is strictly obeyed. This study recommended that spatial heterogeneity should be taken into account in the process of land change modeling and the integrated model can be applied to other areas for further validation and forecasts.     

A land-use and capital-investment allocation optimization model to develop a fair community opportunity framework for Columbus,Ohio

Land-use planning and capital investments have been increasingly recognized as important planning tools to mitigate social inequality. The idea is based on the concept of “neighborhood effects” that the built environment of a city has influence on the socioeconomic outcome of a residential community, which is defined as community opportunity in this study. The explicit local relationships between community opportunity and the physical setting of land uses, transportation infrastructure, and public facilities can be captured by a geographically weighted regression (GWR) model. The GWR results indicate the most effective locations for further developments and investments. This research is one of the few to incorporate such a GWR model into an optimization modeling framework in the contexts of Columbus, Ohio, to (1) maximize the total community opportunities over the region, and (2) minimize the total difference in community opportunities among the 284 census tracts, by an optimal allocation of future land uses and capital investments. Solving the optimization model for the two policy scenarios provides decision makers a new insight into the problem of social inequality. Comparison between the two allocation results has implications for the efforts to improve community opportunity without exacerbating social inequality, by allocating future activities to positive but less effective locations.     

Modelling the relationship between urban expansion processes and urban forest characteristics: An application to the Metropolitan District of Quito

The rapid process of global urbanisation engenders changes in urban socio-ecological systems and in the landscape structure. However, the future processes of urban expansion in Latin American cities has been little studied even though the wellbeing of its citizens will depend on territorial management and on planning the provision of ecosystemic benefits and services. This research, considering different socio-ecological dimensions, proposed to determine the causes of potential urban expansion, analysing the dimensions and possible predictors that would explain the expansion of a high Andean city and its influence on peri-urban forest landscapes.To develop a model that integrates the complexity of the system, we used the following five dimensions: biophysics, land cover and management, infrastructure and services, socio-economics, and landscape metrics, and we opted for a binomial analysis through a spatial logistic regression model developed from 33 predictors.Considering the odd radio of the model, we observe that the independent increase in predictors, including building blocks, drinking water, sewerage, waste collection, average land size, the Interspersion and Juxtaposition Index (IJI) and Largest Patch Index (LPI), and the constant behaviour of the others predictors, would increase the probability of a potential urbanisation of the territory. Similarly, the independent increase in predictors, including the presence of protected areas, the presence of protected forests, land cover, unemployment, and the Shannon Diversity Index(SHDI), reduce the probability of the urbanisation process.Our results suggest that the territorial vulnerability from a potential urbanisation process is strongly related to an increase in infrastructure, services, and the average size of properties variables. Moreover, the landscape with the greatest potential for urbanisation presents an adequate intercalation of the different patches that compose it. However, the presence of variables such as protected areas and protective forests, in addition to monitoring indicators such as landscape diversity and mitigation strategies, could be considered to focus the analysis on the current dynamics of urbanisation processes in Latin America.